arXiv:2607. 20494v1 Announce Type: new Abstract: Production LLM applications commonly stack a regex filter in front of model-side alignment; prior work found no measurable coverage gain from adding a live Gemini backend behind an active regex filter.
arXiv:2606. 02822v1 Announce Type: cross Abstract: Production LLM applications stack several defense families -- refusal-phrase filters, token-budget controls, model allowlists, rate limits, tool-registry authentication -- yet existing breach-and-attack-simulation (BAS) benchmarks report a single aggregate coverage number, hiding which family closes which threat.
arXiv:2607. 01239v1 Announce Type: cross Abstract: Character-level perturbations bypass safety alignment in modern LLMs despite leaving prompts human-readable.
OpenSanctions Pairs is the first large‑scale public benchmark for entity matching on sanctions and OSINT data, comprising 755,540 expert‑labeled pairs drawn from over 1 million entities across 293 source datasets and 45 jurisdictions. The dataset spans multiple languages and writing systems, inconsistent structures, and time‑varying provenance, making it far more heterogeneous than prior benchmarks. Baseline experiments show a rule‑based matcher achieving 91.3 % F1, GPT‑4o reaching 99.0 % F1, and a locally deployable open‑source model scoring 98.2 % F1, with complementary failure modes that highlight the need to focus on downstream pipeline components.
The paper evaluates nine on‑device named‑entity recognition models ranging from classical taggers to large language models, measuring not only accuracy but also latency and output validity. Using a silver‑gold benchmark derived from an LLM judge panel and a human‑validated corpus, the study shows that encoder‑based models achieve comparable accuracy to a 4 B instruct LLM while being much smaller, faster, and producing no malformed output. Confidence calibration of GLiNER is analyzed, revealing over‑confidence but improved reliability after temperature scaling and thresholding.
arXiv:2609. 31181v1 Announce Type: new Abstract: Black-box model identification works by scoring a model's response to natural-language prompts.
arXiv:2605.01699v4 Announce Type: replace Abstract: Recent attacks show that behavioural unlearning of large language models leaves internal traces recoverable by adversarial probes. We characterise...
arXiv:2608. 10216v1 Announce Type: cross Abstract: Agent frameworks ship quality gates that compare text blocks by embedding-cosine similarity and decide at a fixed cutoff.
The paper introduces AgentDiff, a metric that quantifies how much LLM agents’ answers differ when inputs are altered by meaning‑bearing rewrites (paraphrases, synonym substitutions) versus presentation changes (reordering, formatting, distractors). Across 68 model–benchmark–scaffold combinations involving ten LLMs and over 1,500 questions, meaning‑bearing rewrites consistently produce a roughly 20‑percentage‑point higher inconsistency rate than presentation changes, a gap that persists across severity proxies and remains significant even outside the Qwen family. Trace analysis reveals that meaning‑bearing rewrites preserve the first action but reduce thought similarity from the second step onward, extending the divergence cascade—a phenomenon termed “stealth divergence.”
arXiv:2608. 03172v1 Announce Type: new Abstract: Structure-preserving de-identification replaces protected health information (PHI) with realistic same-type surrogates -- "Anna S.
arXiv:2607. 13346v1 Announce Type: cross Abstract: Alignment faking is dangerous because a model can appear compliant under monitoring while preserving behavior it would reveal when unmonitored.
The paper introduces a reference‑based bias detection method that audits hidden‑state representations of language models by encoding sentences as similarities to a fixed set of anchor sentences. This relative representation allows comparison across model variants, such as before and after fine‑tuning, and yields a metric called Representational Bias Shift (ΔB). ΔB correlates strongly with output‑level bias changes, can detect bias‑increasing checkpoints with high ROC AUC, and is computationally efficient, requiring only a few minutes and far less compute than traditional benchmarks.
The paper reports that in English all‑words word sense disambiguation (WSD), the scarcity of high‑quality labels—not the models—has become the limiting factor. The authors introduce lexEN, a human‑adjudicated correction layer over the Maru2022 ALL_NEW benchmark, and SenseBench, a living leaderboard for LLM WSD evaluation. They show that frontier large language models reach about 95 % accuracy on lexEN‑v1, that relabeling corpora with these models improves downstream systems, and that fine‑grained WordNet senses are often ill‑posed, with coarsening improving both annotator agreement and model performance. "whyItMatters":"The study highlights that improving label quality and managing annotation costs are now the critical challenges for advancing WSD performance, as model accuracy is already near its theoretical ceiling."